Voquill: the open source alternative to WisprFlow

WisprFlow is a good product. I've used it. If you just want voice typing on your laptop and you're happy paying a subscription forever, it does the job.

The problem is that it's a closed box. You pay monthly, and you have no real way to know what's happening to your voice once it leaves your machine. For a tool that listens to you all day, that bothered me enough to build an alternative.

Voquill does everything WisprFlow does. The source code is public, you pick where your audio actually goes, and for most people it ends up being free.

Open source, all of it

The desktop app, the mobile apps, the API, the enterprise deployment, all of it is on GitHub. You can read it, build it yourself, or fork it if you want to change something.

Voice is personal in a way text isn't. It's your unfiltered thoughts, the email you're rattling off to your boss, the note from a doctor's appointment, the first draft of whatever you've been putting off. I'd rather you be able to verify what happens to that audio than trust a privacy policy.

Four ways to run it

This is the part WisprFlow can't match.

Run the models locally. Whisper for transcription, a local LLM for the cleanup pass, all on your own machine. No internet required, no audio leaves your device. A decent laptop handles it fine.

Bring your own API key. Groq, OpenAI, Anthropic, about a dozen providers are supported. Your audio goes straight from your machine to whoever you already trust. We never touch it. This path is free forever, you're just paying whatever your provider charges (for Groq right now that's nothing).

Use Voquill Cloud. If you don't want to think about any of the above, we run a managed cloud for $12 a month. It's the closest equivalent to the WisprFlow experience, except the client is still open source and you can leave any time.

Self-host the whole backend. Deploy the Voquill server to your own cloud or an on-prem box, and every piece of audio stays inside your network. Legal, healthcare, finance, government, the people who literally cannot send voice data to a random SaaS, this is the answer.

AI cleanup you control

Raw transcripts are rough. Filler words, weird punctuation, run-ons, the occasional misheard word. Voquill runs the transcript through an LLM afterwards to clean it up.

What changes compared to other tools is that you pick the model. Small local LLM if you want everything offline. A fast hosted one like Groq if you care about speed. Claude or GPT if you want the output to sound genuinely good. It also adapts to context, so an email comes out looking like an email, and code comments stay terse.

Works on every platform

WisprFlow is Mac and Windows. That's it. If you're on Linux, or you want to dictate on your phone, there's nothing for you.

Voquill runs on macOS, Windows, Linux, iOS, and Android. Same account across all of them. Dictate on your laptop during the day, pick up your phone on the train, keep going.

What it actually costs

If you're currently paying for WisprFlow, the free BYO-key path is probably enough on its own. There's a five minute setup guide with Groq if you want to walk through it.

Why this exists

I didn't build Voquill to dunk on WisprFlow. I built it because closed and subscription-only was the only option for voice typing, and that felt wrong for a tool that listens to you all day. You should be able to see what's running, pick where your data goes, and not pay forever if you don't want to.

Download Voquill and try it. If you're coming from WisprFlow you'll be set up in about 2 minutes.

Questions or weird edge cases, hop on our Discord or email support@voquill.com. We read everything.